Zebrafish capable of generating future state prediction error show improved active avoidance behavior in virtual reality

Hitoshi Okamoto,Tomoki Fukai,Makio Torigoe,Hisaya Kakinuma,Tazu Aoki,Chi Chung Alan Fung,Takuya Isomura,Hideaki Shimazaki,Tanvir Islam

doi:10.1038/s41467-021-26010-7

Abstract

Animals make decisions under the principle of reward value maximization and surprise minimization. It is still unclear how these principles are represented in the brain and are reflected in behavior. We addressed this question using a closed-loop virtual reality system to train adult zebrafish for active avoidance. Analysis of the neural activity of the dorsal pallium during training revealed neural ensembles assigning rules to the colors of the surrounding walls. Additionally, one third of fish generated another ensemble that becomes activated only when the real perceived scenery shows discrepancy from the predicted favorable scenery. The fish with the latter ensemble escape more efficiently than the fish with the former ensembles alone, even though both fish have successfully learned to escape, consistent with the hypothesis that the latter ensemble guides zebrafish to take action to minimize this prediction error. Our results suggest that zebrafish can use both principles of goal-directed behavior, but with different behavioral consequences depending on the repertoire of the adopted principles.

Highlights

Animals make decisions under the principle of reward value maximization and surprise minimization
The system alternated between the presentation of the visual stimuli on the four liquid-crystal displays (LCDs) and the imaging of the neural activity by the photomultiplier tube (PMT) of the microscope (Fig. 1e)
We examined whether our observation—i.e., the straight swimming pattern of the fish with the scenery flow prediction error (SFPE) ensemble—could be explained by the hypothesis that the SFPE played a role in facilitating taking optimal action to minimize the prediction error between the real observed scenery and the predicted backward moving scenery

Summary

Introduction

Animals make decisions under the principle of reward value maximization and surprise minimization. To address whether zebrafish are able to use both value maximization and surprise minimization[2,3], we established a closedloop virtual reality two-photon calcium imaging system in which the surrounding scenery moved backward in response to the tail beating of the fish (Fig. 1a) We used this system to study active or passive avoidance behavior as an example of the simple behavioral paradigm of goal-directed behaviors[6,7,8,9]. Non-negative matrix factorization (NMF)[20,21], partitioned the complex activity pattern of the entire neural population into a linear superimposition of the neural activity of multiple elemental ensembles Using this virtual reality system, we found that adult zebrafish can learn both GO (active avoidance) and NOGO (passive avoidance) tasks in a series of trials carried out within one day. Under limited time constraint of the trials, zebrafish using both principles exhibit optimized escape behavior compared with zebrafish behaving with only the reward-based principle

Objectives

Methods

Results

Conclusion